U.S. patent application number 13/045064 was filed with the patent office on 2011-11-10 for multicore fibers and associated structures and techniques.
This patent application is currently assigned to OFS FITEL, LLC. Invention is credited to John M. Fini, Thierry F. Taunay, Man F. Yan, Benyuan Zhu.
Application Number | 20110274398 13/045064 |
Document ID | / |
Family ID | 44563850 |
Filed Date | 2011-11-10 |
United States Patent
Application |
20110274398 |
Kind Code |
A1 |
Fini; John M. ; et
al. |
November 10, 2011 |
MULTICORE FIBERS AND ASSOCIATED STRUCTURES AND TECHNIQUES
Abstract
A multicore fiber comprises a plurality of cores extending along
the length of a fiber body. Each of the cores is surrounded by a
cladding. The plurality of cores and surrounding cladding provide
respective index variations, so as to form a respective plurality
of waveguides for conducting parallel data transmissions from a
first end of the fiber to a second end. The plurality of cores has
a cross-sectional geometry in which the plurality of cores is
configured in a polygonal array, in which at least some of the
cores are positioned at the vertices of the array. The polygonal
array is configured such that neighboring cores in the array are
separated from each other by a distance that is sufficient to
prevent crosstalk therebetween.
Inventors: |
Fini; John M.; (Jersey City,
NJ) ; Taunay; Thierry F.; (Bridgewater, NJ) ;
Yan; Man F.; (Berkeley Heights, NJ) ; Zhu;
Benyuan; (Princeton, NJ) |
Assignee: |
OFS FITEL, LLC
Norcross
GA
|
Family ID: |
44563850 |
Appl. No.: |
13/045064 |
Filed: |
March 10, 2011 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
61312497 |
Mar 10, 2010 |
|
|
|
61314184 |
Mar 16, 2010 |
|
|
|
Current U.S.
Class: |
385/124 ;
385/126 |
Current CPC
Class: |
G02B 6/02042 20130101;
H04J 14/04 20130101; G02B 6/0365 20130101 |
Class at
Publication: |
385/124 ;
385/126 |
International
Class: |
G02B 6/028 20060101
G02B006/028; G02B 6/02 20060101 G02B006/02 |
Claims
1. A multicore fiber, comprising: a plurality of cores extending
along the length of a fiber body, wherein each of the cores is
surrounded by a cladding, and wherein the plurality of cores and
surrounding cladding provide respective index variations, so as to
form a respective plurality of waveguides for conducting parallel
data transmissions from a first end of the fiber to a second end,
wherein the plurality of cores has a cross-sectional geometry in
which the plurality of cores are configured in a polygonal array,
in which at least some of the cores are positioned at the vertices
of the array, and wherein the polygonal array is configured such
that neighboring cores in the array are separated from each other
by a distance that is sufficient to prevent crosstalk
therebetween.
2. The multicore fiber of claim 1, wherein the fiber has an outer
diameter that is substantially equal to that of a standard
single-mode fiber.
3. The multicore fiber of claim 1, wherein the polygonal array
comprises a hexagonal shape, and wherein at least some of the cores
are positioned at the vertices of the hexagonal shape.
4. The multicore fiber of claim 3, wherein the polygonal array
comprises a regular hexagonal shape.
5. The multicore fiber of claim 4, wherein the plurality of cores
comprises six cores, wherein the six cores are positioned at the
vertices of a regular hexagon.
6. The multicore fiber of claim 4, wherein the plurality of cores
comprises seven cores, wherein six cores are positioned at the
vertices of a regular hexagon, and wherein one core is positioned
at the center of the regular hexagon.
7. The multicore fiber of claim 4, wherein the plurality of cores
comprises nineteen cores, wherein the cores are located within
individual rods; and wherein the rods are positioned to provide a
minimal outer circumference.
8. The multicore fiber of claim 6, wherein the plurality of cores
comprises a plurality of single-mode cores.
9. The multicore fiber of claim 8, wherein the plurality of
single-mode cores is configured for single-mode operation in
wavelength regions including 1310 nm and 1490 nm.
10. The multicore fiber of claim 8, wherein the plurality of
single-mode cores is configured for single-mode operation in a
bandwidth window of 1490 to 1620 nm.
11. The multicore fiber of claim 8, wherein each of the seven cores
has a diameter between 6 .mu.m and 10 .mu.m wherein the
core-to-core pitch is between 30 .mu.m and 50 .mu.m, and wherein
the core-clad difference is between 0.004 and 0.010.
12. The multicore fiber of claim 8, wherein each of the seven cores
has a .DELTA.n between 0.004 and 0.008 and is surrounded by a
cladding region having a .DELTA.n between -0.0008 and -0.0040.
13. The multicore fiber of claim 6, further including a trench for
reducing crosstalk.
14. The multicore fiber of claim 6, wherein the plurality of cores
comprises a plurality of graded-index multimode cores.
15. The multicore fiber of claim 14, wherein the plurality of
graded-index multimode cores is configured for single-mode
operation in wavelength regions including 1310 nm and 1490 nm.
16. The multicore fiber of claim 6, wherein the plurality of cores
comprises cores having a diameter between 15 .mu.m and 32 .mu.m and
wherein the core-to-core pitch is between 35 .mu.m and 45
.mu.m.
17. The multicore fiber of claim 6, wherein the core index
difference is between 0.010 and 0.030.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] The present application claims the priority benefit of the
following United States provisional patent applications, which are
owned by the assignee of the present application, and which are
incorporated herein by reference in their entirety:
[0002] U.S. Prov. Pat. App. Ser. No. 61/314,184, filed on Mar. 16,
2010; and
[0003] U.S. Prov. Pat. App. Ser. No. 61/312,497, filed on Mar. 10,
2010.
BACKGROUND OF THE INVENTION
[0004] 1. Field of the Invention
[0005] The present invention relates generally to multicore optical
fiber designs, devices, and applications.
[0006] 2. Background Art
[0007] Passive optical networks (PONs) are now being deployed
worldwide in large numbers for broadband access services. The rapid
growth in data traffic has recently led to an exponentially growing
demand for capacity in access networks. This growing demand has in
turn driven an increasing need for high counts of feeder fibers,
causing congestion problems in duct pipes, and like structures.
Hence, low-cost, high-density cables with high fiber counts are
necessary to construct practical PON systems for future optical
access networks. Similar needs exist for increasing the capacity of
long-haul, backbone networks, as bandwidth continues to grow
unabated while technological solutions for providing such bandwidth
appear to be saturating.
[0008] Multicore fiber (MCF) offers a possible solution for
increasing fiber density, spectral efficiency per fiber, and for
overcoming cable size limitations and duct congestion problems. The
goal of multicore fiber solutions, and spatial division
multiplexing in general, is to increase the bandwidth capacity of a
communication link at a rate greater than the increase in cost of
conventional solutions. In other words, a system which increases
capacity by a factor N using spatial division multiplexing will be
commercially interesting if the cost is significantly less than N
times the cost of deploying conventional single-spatial-mode
solutions.
[0009] Design and fabrication of several types of MCFs have been
reported to address this need for high density while maintaining
properties similar to those of single-core solutions, such as low
loss, low crosstalk and facile connectivity. The crosstalk level,
i.e. the power transferred between the cores, is determined by the
refractive index profiles of the cores and surrounding cladding, as
well as the core-to-core distance and the physical layout of the
fiber (e.g., bends, twists, strains, and the like). The core
density is dictated by the core-to-core distance and geometrical
arrangement of the multiple cores. The index profile, core
geometry, and coating also affect microbend and macrobend loss, as
well as the nonlinear properties of the fiber. Therefore, a
comprehensive design is necessary to optimize overall optical fiber
parameters for MCF. Another important problem is connectivity:
commercial use of MCF requires low-cost reliable splicing and
coupling of signals into and out of the closely-spaced individual
cores.
[0010] In addition, the demand for ever higher capacity data
transmissions has attracted considerable interest in the
development of high-density and high-speed parallel optical data
links for a wide range of applications including interne switches,
servers, future high performance computers and data centers. A
low-crosstalk and low-loss fiber device that enables coupling to
individual cores is important for parallel MCF transmissions.
[0011] In the case of internet switches, the increase of fiber
bandwidth using DWDM technology leads to aggregate bandwidths in
excess of 1 terabit per second (Tb/s). In addition, system size has
increased from single-shelf to multi-rack configurations.
Intrasystem, rack-to-rack interconnections can span a range of
several meters to tens meters. The task of providing and managing
hundreds of individual links using either copper-based or
conventional fiber cables is becoming increasingly challenging.
[0012] In high performance super-computers and data centers,
thousands to tens of thousands of optical links operating at 1 Gb/s
up to 10 Gb/s may be present. The longest distances for
multichannel parallel links in such systems are typically less than
100 m. The key requirements for ensuring successful deployment of
high-density parallel optical data links in that context include
low cost, high density, rapid installation, and low power
consumption. The majority of work to date has focused on
one-dimensional parallel optical data links, which utilize
multimode fiber ribbons with a one-data-channel-per-fiber
arrangement. Such fiber ribbons typically comprise a 1.times.12
linear array of multimode fibers on a 250 .mu.m pitch. However,
such a system configuration is costly, complicated and bulky.
SUMMARY OF THE INVENTION
[0013] An aspect of the invention provides a multicore fiber,
comprising a plurality of cores extending along the length of a
fiber body. Each of the cores is surrounded by a cladding. The
plurality of cores and surrounding cladding provide respective
index variations, so as to form a respective plurality of
waveguides for conducting parallel data transmissions from a first
end of the fiber to a second end. The plurality of cores has a
cross-sectional geometry in which the plurality of cores is
configured in a polygonal array, in which at least some of the
cores are positioned at the vertices of the array. The polygonal
array is configured such that neighboring cores in the array are
separated from each other by a distance that is sufficient to
prevent crosstalk therebetween. The separation may be uniform, or
non-uniform, and the pattern may or may not be polygonal but may
have other forms of symmetry, such as radial.
[0014] Further aspects of the invention provide single-mode and
multimode multicore fibers. In one practice of the invention, the
plurality of cores is arranged in an array with a hexagonal shape,
with individual cores positioned at respective vertices of the
hexagon.
BRIEF DESCRIPTION OF THE DRAWINGS
[0015] FIGS. 1 and 2 show, respectively, cross section and
isometric diagrams of an exemplary multicore fiber according to an
aspect of the invention.
[0016] FIG. 3 shows a table setting forth exemplary specifications
for the multicore fiber shown in FIGS. 1 and 2.
[0017] FIG. 4 shows a cross section diagram of a multicore fiber
according to a further aspect of the invention.
[0018] FIG. 5 shows a graph illustrating attenuation spectra,
measured using a cutback technique, of a sample of the multicore
fiber shown in FIGS. 1 and 2.
[0019] FIG. 6 shows a table comparing attenuation for the center
core and outer cores of the multicore fiber shown in FIGS. 1 and 2
and a standard single-mode fiber.
[0020] FIG. 7 shows a graph illustrating the calculated tunneling
and macrobend losses for a 130 .mu.m clad diameter multicore
fiber.
[0021] FIG. 8 shows a graph illustrating an exemplary modefield
guided by a core.
[0022] FIG. 9 shows a graph illustrating optical power distribution
vs. radius in an exemplary multicore fiber.
[0023] FIG. 10 shows a table setting forth the measured crosstalk,
after 11.3 km, between the six outer cores and the center core in
an exemplary multicore fiber.
[0024] FIG. 11 shows a composite graph illustrating the effect of
increasing the distance between core and a loss-inducing
feature.
[0025] FIG. 12 is a composite graph illustrating how trenches
reduce the tail of an exemplary modefield.
[0026] FIG. 13 shows a diagram illustrating the use of a down-doped
material in a region between the outer cores and coating in an
exemplary multicore fiber.
[0027] FIGS. 14-16 show a series of graphs illustrating a
calculation showing the ability to reduce tunneling loss by
increasing the distance between a core and a coating interface.
[0028] FIG. 17 shows a graph comparing simulations for several
designs, all using the same core rods with the same core size and
shape and core spacing.
[0029] FIGS. 18A and 18B show, respectively, a cross section
photograph and diagram of a graded-index multicore fiber according
to a further aspect of the invention.
[0030] FIG. 19 shows a refractive index profile of the multicore
fiber shown in FIGS. 18A and 1813, which was measured using a
tomographic index profiler.
[0031] FIG. 20 shows a graph illustrating relative power vs. radius
for crosstalk measurements for the multicore fiber shown in FIGS.
18A and 18B.
[0032] FIG. 21 illustrates a schematic diagram of an experimental
setup used to investigate the high-speed parallel transmission
characteristics of the multicore fiber shown in FIGS. 18A-18B and
19.
[0033] FIG. 22 is a graph 220 showing the performance of the center
channel for back-to-back vs. 100 m multicore fiber transmission
with center channel transmitted only, and with all seven channels
transmitted.
[0034] FIG. 23 is a graph showing the BER performance of all seven
channels after 100 m MCF transmission, when all seven channels are
operated simultaneously.
DETAILED DESCRIPTION
[0035] Aspects of the present invention are addressed to multicore
fibers and associated structures and techniques. The present
description is divided into two sections.
[0036] The first section relates to a 7-core single-mode multicore
fiber, as described in U.S. Provisional Patent Application Ser. No.
61/314,184, filed on Mar. 16, 2010. The second section relates to a
7-core graded-index multimode multicore fiber, as described in U.S.
Provisional Patent Application Ser. No. 61/312,497, filed on Mar.
10, 2010. Both of these applications have been assigned to the
assignee of the present application, and incorporated herein by
reference in their entirety.
[0037] The following discussion is organized as follows:
[0038] I. Multicore Fiber I [0039] A. Introduction [0040] B. Fiber
Design
[0041] II. Multicore Fiber II [0042] A. Introduction [0043] B.
Fiber Design [0044] C. High-Speed Parallel Transmission
[0045] III. Conclusion
I. MULTICORE FIBER I
[0046] A. Introduction
[0047] There is described in Section I an exemplary multicore fiber
(MCF), with seven single-mode cores arranged in a hexagonal array.
The MCF is designed and fabricated for construction of
high-density, high-count optical fiber cables, which can be used
in, for example, optical access networks, core networks, and other
applications requiring multichannel parallel transmission at
distances ranging in the 10's and 100's of kilometers, or even
longer. The described MCF can also be used at longer distances.
Depending upon the transmission distance, it may be necessary to
provide amplification. An amplification system for an MCF
transmission system is described in U.S. Prov. Pat. App. Ser. No.
61/314,181, which is owned by the assignee of the present
application, and which is incorporated herein by reference in its
entirety.
[0048] Using the described design, it is possible to construct a
low-crosstalk multicore fiber having a diameter that is
substantially equal to, and compatible with, that of currently used
standard single-mode, single-core fibers. As used herein, the term
"compatible" means that conventional and well-established methods
for cleaving, fusion splicing and connectorizing standard
single-core fibers can also be used with multicore fibers.
Furthermore, such multicore fiber can also be used in conventional
cable design with relatively little modification of cable design.
For a range of applications, fibers with diameter larger and
smaller than 125 .mu.M have been developed and are now standard.
This includes thin fibers with 80 .mu.m glass diameter as well as
fibers as large as 200 .mu.m, or even 400 .mu.m. Thus, it is
possible to incorporate the described fiber into already existing
physical structures, such as duct pipes and the like, used in
current optical fiber transmission links, with similar installation
and maintenance protocols.
[0049] Described herein are the properties of the exemplary MCF,
including crosstalk, attenuation and splice loss characteristics.
Further described in this section are: a low-crosstalk, low-loss
tapered multicore connector (TMC) for coupling individual signals
into and out of the MCF; and a network configuration in which the
described MCF and TMC are used in a passive optical network (PON).
According to a further aspect of the invention, MCF parallel
transmissions are used in a PON to increase fiber density and to
increase the number of optical network end users at the
subscribers' premises.
[0050] By using the described structures and techniques, it has
been possible to demonstrate, in an exemplary PON, simultaneous
transmissions of 1310 nm and 1490 nm digital signals at 2.5 Gb/s
over 11.3-km of 7-core MCF with a split ratio of 1:64. Thus, in the
present example, the described PON can serve a total of 448
end-users at the subscriber premises from a single fiber.
[0051] B. Fiber Design
[0052] FIGS. 1 and 2 show, respectively, cross section and
isometric diagrams of an exemplary MCF 20 according to an aspect of
the invention. MCF 20 comprises a plurality of cores 22a-g
extending along the length of a fiber body 21. Each of the cores is
surrounded by a cladding 24. The plurality of cores 22a-g and
surrounding cladding 24 provide respective index variations, so as
to form a respective plurality of waveguides for conducting
parallel data transmissions from a first end of the fiber to a
second end. The plurality of cores 22a-g has a cross sectional
geometry in which the plurality of cores are configured in a
polygonal array, in which at least some of the cores 22b-g are
positioned at the vertices of a polygon 26. As described below,
neighboring cores in the plurality of cores are separated from each
other by a distance D that is sufficient to maintain an acceptably
low level of crosstalk therebetween. The spacing D may be uniform
and constant within a certain tolerance, or it may vary
intentionally between various pairs of cores. Such variation may be
desired and useful to manage intra-core effects, such as crosstalk
and optical attenuation, as described below.
[0053] MCF 20 can be fabricated, for example, using a
"stack-and-draw" technique. A preform assembly is formed from a
plurality of core rods and filler rods which fill the gaps between
and around the core rods. The preform assembly may be collapsed
into a fully-solid structure (or interstitial holes may remain in
the structure intentionally) and drawn into a multicore fiber,
wherein the core rods correspond to the fiber cores, and the filler
rods correspond to the cladding.
[0054] The core rods are typically doped with suitable
index-modifying elements to achieve a desired refractive index
profile. The filler rods may be fabricated from an undoped
material, such as plain silica (SiO.sub.2). Alternatively, it may
be desired for at least some of the filler rods to be doped to
achieve a different cladding refractive index around one or more of
the cores. Further, the preform assembly may include other
structures, including capillary tubes, substrate tubes, or the
like. The refractive index of each fiber region is expressed as an
index difference .DELTA.n, i.e., the difference between the
region's refractive index and that of undoped cladding.
[0055] FIG. 3 shows a table 30 setting forth exemplary
specifications for MCF 20, which is configured for single-mode
operation in the 1310 nm and 1490 nm regions. MCF 20 comprises
seven cores 22a-g, each having a nominal diameter of 8 .mu.m. A
center core 22a is positioned at the center of six outer cores
22b-g that are arranged at the vertices of a regular hexagon 26,
with a core-to-core pitch of 38 .mu.m. The outer diameter of the
glass cladding 24 is 130 .mu.m. MCF 20 further includes an acrylate
dual coating (not shown), having an outer diameter of 250 .mu.m.
The measured cutoff wavelength for each core is approximately 1200
nm, and modefield diameters (MFD) at 1300 nm and 1490 nm are
approximately 8.3 .mu.m and 9.3 .mu.m respectively.
[0056] It should be noted that, according to a further aspect of
the invention, the fiber may comprise fewer than seven cores, or
more than 7 cores. For example, it would be possible to omit the
center core, in order to provide 6 channels instead of 7. FIG. 4
shows a cross section diagram of a fiber 40 with such a
configuration. Similar to MCF 20 (FIGS. 1-3). MCF 40 includes six
cores 42 positioned at the vertices of a hexagon. However, no
center core is provided. It would also be possible to employ a core
configuration having a different shape, e.g., rectangular.
[0057] Returning to MCF 20 (FIGS. 1-3), each individual core 20a-g
has an index difference .DELTA.n of approximately 0.0046 surrounded
by a cladding region having a .DELTA.n of -0.0012 relative to the
outermost cladding, resulting in a core-clad index difference
n.sub.core-n.sub.clad.apprxeq.0.0058. Note that in this design, the
cladding adjacent to the core comprises a refractive index
"trench." All of the individual MCF cores have the same, or
substantially equal, dispersion and dispersion slope values. At
1490 nm, the dispersion is approximately 10.5 ps/nm-km, and the
dispersion slope is approximately 0.059 ps/nm.sup.2-km. Similarly,
the cores have substantially equal modefield size and shape,
effective nonlinear properties, and propagation constants.
[0058] FIG. 5 shows a graph 50 illustrating the attenuation
spectra, measured using a cutback technique, of a sample of the MCF
having a length of 11.3 km.
[0059] FIG. 6 shows a table 60 comparing the attenuation for the
center core 24, the outer cores 22, and a standard single-mode
fiber (SSMF).
[0060] As shown in table 60, at 1310 nm, the center core 22 has a
measured attenuation of 0.39 dB/km. At 1490 nm, the center core has
a measured attenuation of 0.30 dB/km. At 1310 nm, the average loss
for the 6 outer cores 24 is 0.41 dB/km. At 1490 nm, the average
loss is 0.53 dB/km.
[0061] As further shown in table 60, the center core loss at 1310
nm and 1490 nm and the outer core loss at the shorter wavelength
(i.e., 1310 nm) may be considered comparable to the loss in a
conventional standard single-mode fiber (SSMF) in that the spectral
shape of the loss curve has well-known features, including: a
hydroxyl overtone around 1380 nm; a loss component which decreases
with wavelength and which may be attributed to Rayleigh scattering;
and an approximately wavelength independent component which may be
attributed to scattering and contamination. As described below,
this latter component may be reduced using improved fiber design
and fabrication methods. At 1310 nm and 1490 nm, the losses of
conventional SSMF are approximately 0.35 dB/km and 0.24 dB/km,
respectively. A notable feature of FIG. 5 is that the MCF outer
core loss at 1490 nm is higher than the loss in the other cores.
There are a number of reasons for this higher loss, including
microbend loss and tunneling loss interactions with the coating at
close proximity, e.g., core-coating effective index matching. The
outer core loss at the longer wavelength can be reduced in a number
of ways, as discussed in greater detail below.
[0062] The tunneling loss was calculated using a two-dimensional
finite difference vector model solver using a perfectly matched
non-reflecting boundary layer. This tunneling loss introduces an
excess attenuation in the outer cores above that in the center
core.
[0063] FIG. 7 shows a graph 70 illustrating the calculated
tunneling and macrobend losses (with the fiber bent in a 15 cm
diameter coil) for the 130 .mu.m clad diameter fiber (plot 71) in
which the outer core centers are 27 .mu.m from the coating. The
reduction in loss is comparable to the measured excess loss. Thus,
FIG. 7 indicates that the tunneling loss can be reduced to a
negligible level by increasing the fiber clad diameter to 140 .mu.m
(plot 72), such that the outer core centers will be .about.32 .mu.m
from the coating, while keeping the same 38 .mu.m core pitch.
[0064] The mechanism responsible for tunneling loss is coupling of
the signal in the outer cores to cladding and coating modes. In
general, one main strategy for reducing excess long-wavelength loss
is to reduce the modefield present at some feature relevant to the
loss mechanism. For example, as the evanescent tail and optical
field strength are made smaller at the feature, there is a
reduction of absorption and scattering due to overlap of that tail
with any absorbing or scattering materials (such as the polymer
coating). There may also be a reduction in coupling between the
core modes and modes supported by the cladding or coating. Because
perturbations to the fiber such as bending, twisting and heating
are known to modify the effective refractive index profile, the
treatment of the overlap of the modefield distribution with fiber
features must also consider such perturbations. In specific
examples, loss may be calculated for a specific perturbation, such
as bending, but mode-field confining strategies that reduce such
mechanisms typically also reduce other sources of loss, such as
absorption and scattering.
[0065] Other strategies for controlling optical attenuation may be
invoked, depending on the loss mechanism. For example, enforcing
limitations on bending of the fiber, such as limiting the minimum
bend radius experienced in the cable in a deployed fiber, or
reducing the refractive index of the polymer coating or some
cladding feature can control the characteristics of the modefield
distribution. A thicker coating or improved mechanical properties
of the coating which reduce stress on the fiber can reduce
microbending loss.
[0066] FIG. 8 shows a graph 80, illustrating that the modefield
(trace 81) guided by a core 82 has an evanescent tail penetrating
into the cladding (stippled area 83). Since the modefield
distribution typically has wider radial extent at longer
wavelengths, losses due to tunneling often increase at longer
wavelength. This is observed in FIG. 6, discussed above.
[0067] An additional source of optical attenuation arises from
contamination introduced during the core rod and fiber fabrication
process. Not shown in FIG. 8 is the interface between the core rod
and the cladding materials used to assemble the MCF preform in a
stack-and-draw process. Because the optical modefield overlaps with
these interfaces, contamination can induce absorption and
scattering.
[0068] Optical crosstalk between adjacent cores is a significant
issue in the design of a multicore fiber. Crosstalk is strongly
influenced by the spatial distribution of the optical modefields
guided by the cores. The optical crosstalk from the center core to
adjacent outer cores in the exemplary MCF 20 was measured by using
1 meter of SSMF to scan the optical power intensity distributions
at the output endface of the fiber.
[0069] A MCF with 11.3-km length was wound onto a 28 cm-diameter
spool, and the center core 22 was spliced to 1-meter of SSMF
launching 1310 nm and 1490 nm signals. The amount of crosstalk was
determined from the ratio of optical power detected at the 6 outer
cores to the optical power detected at the center core the end of
the 11.3 km MCF.
[0070] FIG. 9 shows a graph 90 illustrating optical power
distribution vs. radius in an exemplary MCF. The local maximum
crosstalk occurs at about 38 .mu.m radial position, i.e., at the
center of the adjacent core.
[0071] FIG. 10 shows a table 100 setting forth the measured
crosstalk between the six outer cores and the center core after
11.3 km. Maximum crosstalk is less than -38 dB at 1310 nm, and less
than -24 dB at 1490 nm, consistent with the expected increased
evanescent penetration through the cladding at longer wavelengths,
where the mode effective index is smaller, and MFD is larger.
[0072] It should be noted that, when all seven cores carry signals
simultaneously, the worst-case crosstalk, compared with the case of
signal transmission through one core, would be 6.times. for the
center core and 3.times. for the outer cores. The 6.times. and
3.times. multipliers are based upon the number of immediately
adjacent cores. It should also be noted that the crosstalk
characteristics of a given MCF depend not only on fiber design
(e.g., index profile, core diameter, core-pitch and the like) but
also on fiber length and the layout (e.g., bends, twists, and the
like) along a given optical link.
[0073] Better confinement of the modefield to reduce attenuation
and crosstalk can be implemented in several ways: The distance
(d.sub.core-feature) between the center core and the relevant
feature (other core, polymer, core rod interface etc) can be
increased. For example, losses due to the coating can be reduced my
making the coating farther from the core, or the radius of the core
rod relative to the core radius can be increased, or the
core-to-core spacing can be increased. Alternatively, the index
profile of the core and cladding can be arranged to provide tighter
confinement of the core, for example, by raising the index in all
or part of the core, lowering the index in all or part of the
cladding (for example adding a trench), or both.
[0074] Improving mode confinement typically makes the modefield
diameter smaller, and this is an important tradeoff. While fibers
with modefield diameters larger than that of a standard single-mode
fiber may have other advantages (e.g., low nonlinearity, lower
connection losses), fibers with relatively small modefield
diameters may be desirable for reducing long-wavelength losses and
increasing the packing density of the cores.
[0075] In addition to modefield diameter, an important metric of
mode confinement is the parameter
.kappa.=(n.sub.eff.sup.2-n.sub.clad.sup.2).sup.1/2 where n.sub.eff
is the effective index of a mode and n.sub.clad is the index of the
cladding. Intensity in the tail falls with position x from the core
center as e.sup.-2.kappa.x, and so fibers with a large .kappa. will
show reduced long-wavelength loss.
[0076] FIG. 11 shows a composite graph 110 illustrating the effect
that increasing the distance between core and a loss-inducing
feature reduces the modefield at the feature. The reduction of
evanescent fields needed for reducing excess loss is similar to the
reduction of evanescent fields accomplished in many low-bend-loss,
or so-called bend insensitive, designs. Thus the use of a low-index
annular trench in the cladding surrounding a core may be highly
desirable, and would entail a tradeoff between cutoff, mode size
(MFD or A.sub.eff), and long-wavelength loss very similar to the
tradeoff between cutoff, mode size, and bend loss in current
bend-insensitive fibers such as those conforming to ITU
specification G.657.
[0077] Desirable index profiles for an outer core of a multicore
fiber would include the inner portions (excluding the outer trench)
of the fibers, or fibers generally with a trench region with index
below around -7.times.10.sup.-3 relative to the cladding index.
[0078] FIG. 12 is a composite graph illustrating how trenches
reduce the tail of the modefield. While the long-wavelength loss is
in many ways analogous to the bend loss of a single-core fiber, or
the crosstalk of a multicore fiber (all roughly proportional to the
strength of the evanescent tail of the guided mode at some relevant
location outside the core), the long-wavelength loss is different
in that it primarily impacts the outermost cores of a multicore
fiber.
[0079] Thus several approaches may be desirable to control the
distribution of the optical modefield: [0080] Use refractive index
profiles which provide greater optical confinement for the cores
with greatest overlap with undesirable features, possibly at the
expense of higher cutoff, higher cost, or smaller mode size for
these cores. [0081] Using lower-index materials in the tail region
of the modefield, for example in between the assembled core
elements. [0082] Use an additional low-index layer between the
outer glass surface of the fiber (the coating interface). For
example, this could be a single annular down-doped region or an
air-clad region surrounding all of the cores. This should be
balanced with multi-path interference and other impairments related
to guiding of unwanted modes or supermodes in the cladding.
[0083] FIG. 13 shows a diagram illustrating the use of a down-doped
material in region D between the outer cores and the coating E. Low
index D will effectively truncate the modefield distribution,
reducing overlap with the coating or material which surrounds
region D.
[0084] FIGS. 14-16 are a series of graphs 140, 150, 160,
illustrating a calculation showing the ability to reduce tunneling
loss by increasing the distance between a core and the coating
interface. Also shown is the dependence on the relative index
between the coating and the glass cladding. In this example,
coating index less than around 0.002 above the cladding index
results in effectively no tunneling (curves not visible where
losses are off scale). Therefore, use of a coating with appropriate
index below or approximately equal to the cladding index prevents
excess loss. Use of a coating with sufficiently high index (e.g.
approximately equal to the cladding index) will further reduce
impairments due to unwanted modes by causing their preferential
attenuation, which may be desirable. Note that treatment of the
coating index should also include temperature effects which may be
significant for practical fiber and cable installations due to the
high value of dn/dT for polymers relative to silica-based
materials.
[0085] Further simulations for a structure with two or three
different cladding index values are illustrated in FIGS. 14-16.
[0086] FIG. 17 shows a graph 170, comparing simulations for several
designs, all using the same core rods (regions A and B in FIG. 13,
standard SMF) with the same core size and shape and core spacing
(40 .mu.m). The intermediate region between core rods, C, is
slightly down-doped (.DELTA.n.about.-0.0012 relative to pure
silica) with a first F320 overclad of each core rod to reduce
evanescent field tails beyond the material of the core rod.
[0087] The FIG. 17 graph 170 illustrates the effect of the index
and thickness of region D in FIG. 13. Large thickness but matched
index between regions C and D drastically reduces tunneling loss,
but results in larger fiber outer diameter. Modest index reduction
of region D gives modest loss reduction.
II. MULTICORE FIBER II
[0088] A. Introduction
[0089] There is described in Section II a graded-index multicore
fiber and related structures and techniques for use in the types of
long or medium distance applications described above as well as
super computers, data centers, and other applications requiring
high-speed parallel transmissions at shorter transmission lengths
(e.g., on the order of 100 m).
[0090] One of the promising solutions for high density parallel
optical data links is to use the multicore fiber (MCF) design
described hereinbelow, which is butt-coupled with 2-dimensional.
VCSEL and PiN photo detector arrays to realize MCF transmissions. A
two-channel simultaneous transmission over a 2.times.2 MCF using
direct coupling with a linear VCSEL array at <1-Gb/s has been
reported.
[0091] For the purposes of laboratory testing, due to the cost of
2-D VCSEL arrays, a parallel high-speed MCF transmission was
conducted using tapered multicore connectors (TMC) for coupling of
the signals into and out of a sample MCF. Commercially available
VCSELs were used as the transmitters, and commercially available
PiNs were used as the receivers. As described below, using the
described structures and techniques, it is possible to demonstrate
10-Gb/s per channel (core) simultaneous transmissions in all seven
cores over 100 meter using tapered multicore connectors and 850-nm
VCSELs, thereby demonstrating a high-speed multicore fiber
transmission for parallel optical data links.
[0092] B. Fiber Design
[0093] FIGS. 18A and 18B show, respectively, a cross section
photograph and diagram of an MCF 180 according to an aspect of the
invention. The MCF comprises seven cores arranged in a hexagonal
array including a center core 181 and six outer cores 182
positioned at the vertices of a regular hexagon 183 in a common
cladding 184. The diameter of each of the cores is 26 .mu.m, and
the core pitch is 39 .mu.m. The cladding diameter is 125 .mu.m and
the acrylate dual coating layer (not shown) is 250 .mu.m. These
diameters are compatible with the diameters of conventional fiber.
According to a practice of the invention, the described MCF is
manufactured using a stack-and-draw process.
[0094] To achieve high-speed (>10 Gb/s) parallel multi-core
fiber transmission, it is necessary for there to be little or no
crosstalk between individual MCF cores to minimize interference
between the distinct signals transmitted through each distinct
core. In addition, an optimized design of a suitable graded index
profile in each core, as well as precise control during the
fabrication process, are essential to maintain uncompromised modal
bandwidth, to precisely control the properties of the low-order and
high-order modes, and to minimize or eliminate deformation.
Furthermore, it is imperative that the refractive index profile and
core rod properties be sufficiently robust that the MCF can be
manufactured with high yield.
[0095] Even though the cores support multiple modes, the same
design considerations as discussed above for single-mode cores may
be employed for controlling crosstalk, spliceability and optical
attenuation. However, the treatment must extend to all modes
supported by the cores, or at least to the most problematic
modes.
[0096] FIG. 19 shows a refractive index profile 190 of MCF 180,
which was measured using a tomographic index profiler. The core
index difference .DELTA.n is approximately 0.016. It can be seen
from FIG. 19 that all seven cores are fairly circular and maintain
an optimized graded-index profile design. The average transmission
loss for the seven cores at 1310 nm is approximately 0.5 dB/km,
which is typical of conventional graded-index fiber.
[0097] Optical crosstalk between adjacent signal-transmitting cores
is an important issue in MCF transmission systems. The crosstalk
characteristics of an MCF not only depend on fiber design
considerations, such as index profile, geometric structure, and the
like, but also on fiber length, bending properties, and the like.
Optical crosstalk from the center core to adjacent outer cores can
be measured by scanning the optical power intensity distributions
at the output endface of the fiber.
[0098] In one test, a 550 m length of fiber was wound onto a spool
having a diameter of 17 cm, and the center core was excited by a
multimode VCSEL operating at a wavelength of 850 nm. FIG. 20 is a
graph 200 illustrating relative power (P.sub.(r))/P.sub.c) vs.
radius for crosstalk measurements, where P.sub.c is the power
measured at center core. The crosstalk of the six outer cores from
the center core in the 550 m fiber were all measured to be all
below -40 dB.
[0099] C. High-Speed Parallel Transmission
[0100] High-speed parallel transmission characteristics of the MCF
are investigated by using a tapered mode coupler (TMC), which is
used to couple the individual signals into and out of an MCF.
[0101] FIG. 21 illustrates a schematic diagram of an experimental
setup 210 which was used to investigate the high-speed parallel
transmission characteristics of a multicore fiber of the type
described above. Setup 210 comprises the following components:
[0102] a 100 m length of MCF 211;
[0103] a first tapered multicore coupler 212 that is fusion-spliced
to an upstream end of MCF 211;
[0104] a second tapered multicore coupler 213 that is
fusion-spliced to a downstream end of MCF 211;
[0105] a transmitter optical subassembly (TOSAs) 214, comprising a
plurality of individual vertical-cavity surface emitting lasers
(VCSELs) connected to respective fiber leads into the first TMC
212; and
[0106] a receiver optical subassemblies (ROSAs) 215, comprising a
plurality of individual PiN detectors connected to respective leads
out of the second TMC 213.
[0107] The core diameter and core pitch of the first and second
TMCs 212, 213 are configured to match those of the MCF 211.
Crosstalk between cores in TMC device is required to be below -25
dB. Also, insertion loss from the TMC must be as small as
possible.
[0108] In one experiment, seven commercially available 850-nm 10
Gb/s VCSELs (TOSA) 214, pigtailed with 1 meter standard 50 .mu.m
multimode fiber (MMF), were connected to each pigtail fiber of the
first TMC 212. All 7 channels operated at 10-Gbps with a
non-return-to-zero (NRZ) 231-1 PRBS signal; the electrical 10-Gb/s
signals were generated from a pattern generator, amplified, divided
with different lengths of microwave cables, and fed into the VCSELs
drivers. The averaged (modulated) optical power was approximately
-0.5 dBm, the laser RMS line-width was approximately 0.35 nm, and
the laser relative intensity noise (RIN) was -125 dB/Hz. For low
cost, commercially available 850-nm GsAs PIN (ROSA) receivers 215
without clock and data recovery (CDR) circuits were used as the
receivers with electrical bandwidth of about 7.5 GHz, and were
connected by means of an approximately 1-meter segment of standard
50 .mu.m MMF to each pigtail fiber of the second TMC 213.
[0109] For back-to-back transmission, the VCSEL transmitter was
connected to the receiver using two 1-meter segments of standard 50
.mu.m MMF and a tunable optical attenuator (matched to 50 .mu.m
multimode fiber). The receiver sensitivity (BER at 10.sup.-12) was
approximately -6.5 dBm, and the rise and fall times (20%-80%) were
typically 51.1 ps and 45.3 ps. The RMS time jitter was
approximately 6.04 ps.
[0110] FIG. 22 is a graph 220 showing the performance of the center
channel for back-to-back vs. 100 m multi-core fiber transmission
with center channel transmitted only, and with all 7 channels
transmitted. The eye diagrams at back-to-back and after 100-m MCF
transmission are shown, respectively, in insets 220a and 220b.
[0111] After 100 m MCF transmission, the rise and fall times were
typically 52.4 and 49.1 ps, and the RMS time jitter was
approximately 6.11 ps. The optical power penalty after MCF
transmission was approximately 0.35 dB. It can be seen in FIG. 22
that there was virtually no penalty (within measurement accuracy)
when all 7 channels operated simultaneously. This means that the
crosstalk between the outer cores and the center core was very low,
which is consistent with the data shown in FIG. 19. The relative
high receiver sensitivity at back-to-back is due to the low-quality
TOSA (VCSELs) and ROSA used, and the low-quality 10-Gb/s electrical
signals resulting from the use of numerous amplifiers and divider
circuits. Nevertheless, error-free operation was achieved in the
experiment.
[0112] FIG. 23 is a graph 230 showing the BER performance of all 7
channels (cores) when all 7 channels are operated simultaneously,
after 100 m MCF transmission. All 7 channels have similar BER
performance with the similar receiver power sensitivity, within
measurement accuracy. This result is consistent with the
index-profile measurement data shown in FIG. 19, that all cores
have similar index-profile. The averaged loss budget of the links,
taken from the 7 channels, is approximately 2.8 dB. The majority of
loss originates from the two TFB due to fabrication process
imperfections. The budget is large enough to realize 7.times.10
Gb-ps parallel optical links.
[0113] It should be noted that the 10 Gb/s per core MCF
transmission demonstrated in the experiment is limited by the speed
of VCSEL and PIN detectors used in the experiment. The relative
small core diameter (26 .mu.m core diameter in MCF compared with
50/125 .mu.m OM3 single-core multimode fiber), which means fewer
optical modes in each core, and well-defined index profiles,
indicate large bandwidth and small modal noise, potentially for
high speed and longer distance transmission.
III. CONCLUSION
[0114] While the foregoing description includes details which will
enable those skilled in the art to practice the invention, it
should be recognized that the description is illustrative in nature
and that many modifications and variations thereof will be apparent
to those skilled in the art having the benefit of these teachings.
It is accordingly intended that the invention herein be defined
solely by the claims appended hereto and that the claims be
interpreted as broadly as permitted by the prior art.
* * * * *